Despite the fact the two K10 Kuma CPUs were running at 1,066MHz DDR2, 5-5-5-15-2T versus 800MHz 4-4-4-12-2-T on the older K8 Brisbane CPUs, the clock to clock raw memory performance isn't hugely impressive as the new 2.7GHz Athlon X2 7750 performs notably slower than the 2.7GHz 5200+ in Everest. The gap is closer at 2.5GHz between the 7550 and 4850e, in favour of the 7550 slightly, however it's mostly due to the fact that the K10 memory controller is not only at a lower frequency (1.8GHz on K10 versus "CPU core frequency" on K8) but accessing it is also a higher latency. Both these negate the effect of using faster DDR2 memory.
All the AMD CPUs are faster than the Intel E5200 though by a considerable margin, and their integrated memory controller also makes the round trip to memory ~36 to 48 percent faster for the AMDs in a clock to clock comparison too. As you can see though, we hit a strange bug in Everest that we couldn't seem to rectify for the E5200. Our E8500 worked fine, but even older 4.5 or 4.2 versions of Everest would only give a 0MB/s copy performance reading.